Dennis Feucht, Innovatia Laboratories
EDN
What makes a legacy IC? Here we take a look at some original parts and discuss why it might be a good idea to revisit their use.
As electronics technology matures, it shows some signs of aging. As innovation wanes, breakthroughs grow farther apart in time, and risktaking decreases. In earlier decades, some IC companies were willing to put on the market a few unusual and conceptually original parts. Signetics, for example, dared in the early 1970s to produce a novel IC – Hans Camenzind’s 555 timer – and it has become a premier legacy IC.
What about other circuits? This article recounts a few of these parts and why it might be good to bring them back.
AD639 sine converter
Long ago, Barrie Gilbert did some investigative work on the two-bipolar-junction-transistor (BJT) differential-amplifier circuit, which has a hyperbolic tangent, tanh, transfer function:
where
I0 is the emitter source current,
VT = 26 mV – the thermal voltage.
To reduce nonlinearity, external emitter resistance, RE, is placed in series with the emitter in amplifiers. Gilbert, however, applied the engineering adage that “if you can’t fix it, feature it” and put it to good use.
Hyperbolic tangents are vaguely related to trigonometric functions, and Alan Grebene at Exar used a single diff amp in the XR2206 function- generator (FG) IC to convert a triangle wave into a sine wave. The result was somewhat acceptable for a first-generation effort, though Gilbert carried out more refinement of the basic idea. He developed the multitanh concept of adding the outputs of diff amps that had inputs offset from each other by a fixed voltage. This approach extended the function (and input range) and also led to other novelties, including the one that is used in the AD639 sine converter.
This IC is a trigonometric wonderland in 16 pins, and because its functional capability is so powerful, it was destined to become a legacy IC. Alas, ADI pulled the AD639 from the market without a replacement. I do not know why. Even Gilbert did not know why. It seemed like a part destined to become a legend. It can synthesize all of the basic trigonometric functions (sin, cos, tan sec, csc, cot) and their inverses.
The sine function is accurate to 0.02%, better than most FG sine outputs and better than the total harmonic distortion (THD) of many audio amplifiers. The IC has two of them, plus offset circuitry, and a multiplier and divider. It was given niche-market pricing and thus did not find its way into FG instruments and others requiring accurate or low-THD sinewave generation. It is specified to 1.5 MHz.
Perhaps the only problem was that because the AD639 was so appealing, ADI put a premium price tag on it, which defeated its spread into the market as a commodity part. Perhaps Rochester Electronics – the leading supplier on the “trailing edge” – could revive it and reap the small fortune that it was destined to produce. There is no reason to confine the mission of Rochester Electronics to that of a replacement-parts supplier for obsolete equipment when it could also be considered the fulfiller of the destiny of great parts for new designs that did not catch on the first time around.
CA3096 BJT array
Somewhat like the AD639 sine converters are the highly versatile building blocks that are transistor arrays. RCA came out with some BJT arrays comprising a CA3000-series line. Some of these parts had fTs of over 1 GHz for NPN BJTs, which make them good for new designs today. When RCA was shuffled around, it eventually ended up with Intersil but had lost the fab facility with its old large-geometry process. Tektronix based the design of the vertical amplifier in its 2205 oscilloscope on CA3046 (or the equivalent National LM3046) parts, which are attractive for implementing fast two- or four-quadrant multipliers (Figure 1).
 |
|
| Figure 1. | Tektronix 2205 Portable Oscilloscope, 20 MHz, 2 Channel. |
Intersil inherited a large but finite supply of these parts that is still available but dwindling. They ought to be brought back on an existing process. It is not a major development project, and such parts would be extremely useful. Intersil did create a replacement in the form of HFA3000-series SOIC parts with multiple-gigahertz fTs but with correspondingly lower breakdown voltages (Figure 2).
 |
|
| Figure 2. | Replacements for the CA3000 series in the form of HFA3000-series SOIC parts offer multiple-gigahertz fTs but correspondingly lower breakdown voltages. The table shown is excerpted from Intersil parts data. |
| Table 1. Differential Pair Matching Characteristics for the HFA3046. | ||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||
While the original CA3000 series was good for ±12 V supplies, the HFA series is better designed with ±5 V supplies, though the ICs can handle up to about 10 V. What is much improved in the HFA series are the PNP BJTs, which are dielectrically isolated rather than made as lateral transistors, as in the CA3096 (Figure 3).
 |
|
| Figure 3. | The CA3096’s BJTs are constructed as lateral transistors. |
The CA3096 is a versatile part, with three NPN and two PNP BJTs. The one drawback is that the lateral PNPs have an fT of only about 6 MHz. (Making the base thin is difficult for lateral BJTs.) For numerous circuits, however, this specification is not a major impediment.
In one example, a feedback amplifier has a quasistatic gain of 3 and a bandwidth of over 50 MHz (Figure 4). It has two forward paths, the slow path through the PNP current mirror and the fast path through Q2 of the diff-amp input stage. It uses all five array BJTs. The only other semiconductor part is the avalanche diode, Z1.
 |
|
| Figure 4. | This feedback amplifier has a quasistatic gain of 3 and a bandwidth of over 50 MHz. It uses all five of the CA3096’s array BJTs; the only other semiconductor part is the avalanche diode, Z1. |
You would not be inclined to design this circuit into a new product because of an uncertain component supply. In addition, the HFA part does not have the voltage range. A series comparable to the CA3000 but with dielectrically isolated PNPs would be a welcomed addition to the neo-legacy category of ICs.
MC14500B industrial-control unit
The 16-pin MC14500B from Motorola is a single-bit, 1-MHz CMOS processor. It has three single-bit registers (flops) and an arithmetic logic unit and executes 16 instructions. Newer microcontrollers blow this part away, but that is not the point. It is a chunk of versatile logic that needs only an external counter for a program counter (PC) and a program memory driven by the PC.
The data memory is also the I/O memory. Four bits of the memory output drive the op-code input on the MC14500B; the rest are I/O addressing of 8-bit bidirectional latches (MC14599B) and 8-input multiplexers or data selectors (MC14512).
The one-bit accumulator is called the result register (RR). Instructions include Load RR, Load the complement of RR, AND data with RR, Complement data and AND, OR, Complement data and OR, exclusive NOR (equivalence), store and store complement pulse the write line with valid RR output, move input data to input register or to output register, skip next instruction if RR=0, and pulse flag O out or flag F out. Two other instructions, JMP and RTN, also output flag pulses. The JMP flag can be used to load an address into the PC. The RTN instruction outputs an RTN flag and skips the next instruction.
The built-in oscillator generates the clock that drives the PC. The rising edge of the clock increments the PC, and while high, the instruction is fetched. During the low phase of the clock, the instruction is decoded and executed. Because it uses bit-serial processing and is I/O intensive, what advantage does this part have nowadays? With the requirement of an additional counter, and program and data memory, this part will remain obsolete because it cannot compete with lower-cost 8- to 16-pin flash-programmable ICs that are easy to use and are far more powerful. Although the part is interesting, it requires too much bit-twiddling to be welcomed back into production. This one will remain forgotten, despite its inspiration quotient.
MC14549 and MC14559 SARs
These successive-approximation registers (SARs) were originally part of the Motorola 4000-series offering of CMOS (mostly) digital ICs. They have 8 bits per IC and can be cascaded for more bits. They are used to build successive-approximation ADCs. Internally, they have a shift register and a parallel-loading register.
Despite the simplicity of an SAR, it is a useful digital function. The SA algorithm searches a range by making a succession of boolean comparisons beginning at midrange. If the voltage is greater, the most-significant bit is set and the next bit is tested until all bits have been determined. All conversions on n bits take n clock cycles, independent of the digitized value.
With an extra comparator and one or two SAR ICs driving an extra DAC, a simple ADC can be added to a system with leftover subparts. Although this level of integration is semi-discrete nowadays, for many applications with multi- DAC and multicomparator ICs – and with a need for a simple ADC – it can be a design possibility.
SARs also could be used for auto-ranging, with fewer average steps than sequential ranging. Similarly, the gain of a variable-gain amplifier (VGA) can be set by an SA search through the large range of the VGA. The weighting of the bits might no longer be binary, but instead be a decade or 1-2-5 sequence. If it’s monotonic, however, the scheme works.
MC4530 dual five-input majority gate
One of the stranger logic functions to make it into integrated form and onto the market was a dual five-input majority gate, sold by Motorola. If three or more of the five inputs are asserted, the output is asserted. Those of you who like to find novel uses for existing logic parts might have an intriguing time with this part. The output was gated with an XNOR (equivalence) gate from a W input, to set the polarity of the output.
What are its uses? This part is relevant to unusual applications, but it can coax some creative thinking. It generates a decision when five or fewer subsystems indicate a status in a redundant system. If, for example, the vital-sign monitors in a hospital’s intensive-care unit show three or more of five patients are in trouble, a triage state is asserted.
By connecting one input high and one low, two out of three control computers (as in the space shuttle) prevail in a vote on an output assertion. Alternatively, if multiple banks of capacitors are asynchronously charged, sufficient charge is available and the firing device is enabled if m out of n banks indicate that they are charged. The function is egalitarian; any m out of n will trigger an event. Using inverted input logic, a statically stable, multi-ped robot with fewer than m of n feet on the ground triggers a fault state.
A voting hierarchy is implemented by cascading the output of one of the majority gates into one input of another. The first five then have one vote in the second five. Although such uses might be practical, it is still an unusual logic function to try to apply. No wonder it is a forgotten circuit.
